Or judging the depth and subtlety of ones' models. For example, I know all of these things (transformer design, bulk core properties), but I can't tell you why they happen. I'd love to rattle off the equations, if there were any.
This is ultimately rooted in the incredible complexity of condensed matter physics, where even tiny impurities, vacancies, dislocations, defects in the magnetic crystal can have profound, unpredictable effects on its macroscopic properties.
It's an empirical model. We note (but arguably not really
understand) parameters like initial and average permeability, saturation, core loss and so on. For which, you merely need to perform dozens or hundreds of measurements, then fit curves, and there are your parameters. For that one core, in that size, in that condition...
A more basic theoretical model is in principle possible, but no closed-form solution is possible (that would produce our empirical parameters from physical inputs), and no approximate solution is known. And so we also come to the economy of theory itself: it would require decades of intensive work to derive such solutions, but the economic value in those solutions is minuscule, so no one will solve them. At least until such time as that effort can clearly be amortized over a large enough time period, or divided into small enough pieces, or -- in my opinion more likely -- solved as a side-effect of some more important condensed-matter physics problem (say, nanotechnology for various purposes: computation, data storage, chemical processing, bioengineering, etc.).
The applied outcome of this, to this thread, would be: if you knew the core material and its exact condition, you could in principle know its nonlinear parameters, and perhaps not by measuring a single point on each winding, but several excitation values for each, or at different frequencies, or even waveforms say -- you would be able to match them up, and determine the winding ratio with good confidence. But clearly this is much more difficult than, say, even x-raying the transformer to count its turns directly -- so "basically no" is a good answer.
Tim